Changing Traffic Patterns in Washington, DC: Fallout from the COVID-19 Pandemic

PPOL 6805: GIS for Spatial Data Science Final Project

Author
Affiliation

Holt Cochran

Georgetown University

Introduction

Washington, DC is frequently listed as having some of worst traffic in the United States (1). There are several factors that contribute to the large amount of traffic in the District: the large and dense population within the city limits packs people into a small land area; bottlenecks exist in certain areas of the city, such as the bridges crossing the Potomac River; drivers commuting from the sprawling suburban areas in Virginia and Maryland are funneled into tight corridors of the city; peak hours of the day, when people commute to and from work, escalate the traffic. Taken together, the District of Columbia experiences high rates of traffic, however there is no one single factor that contributes to the large build-up of drivers within the city (2) - rather, it is a confluence of characteristics of the city and drivers that contribute to vehicle backups.

When COVID-19 began spreading rapidly across the United States in March 2020, businesses, schools, and establishments closed to limit interactions and spread of the virus (3). People stayed home to shelter, and many functions of society were taken online in a remote setting (4). This caused an almost immediate stop in the number of people commuting into cities, including Washington, DC, for work, school, or other functions. The ability to work, learn, and exist in a remote capacity removed reasons for continuing to live in the same areas as before the pandemic - research shows that during the pandemic, large swaths of the US population relocated or moved to other areas (5).

Since the height of the COVID-19 lockdowns, society has reopened to in-person interactions, but remote functions of work, school, doctors visits, etc. still exist. My research for this project centers on how changes in human behavior during the COVID-19 pandemic have affected traffic patterns in Washington, DC and what city policymakers should do to address changes in traffic trends.

Hypothesis: Traffic has increased in Washington, DC since the reopening of schools, businesses, and return to office policies, but not to the same level as before the pandemic. However, do to mass relocation of people during the pandemic, traffic patterns and trends have changed in Washington, DC since COVID-19. Traffic has to become more spread out across the city, as people have moved and traffic overall has decreased, leading to new patterns in areas that did not previously have bad traffic. City policies need adjusting to address these changes and alleviate new traffic patterns and trends.

Literature Review

Methodology

To research changes in traffic patterns in Washington, DC, I am using data from District of Columbia Department of Transportation (DDOT) from 2018 - 2022 (6). The data tracks traffic at a yearly level - these are reported in a statistic called Annual Average Daily Traffic (AADT), which is the average daily traffic a road experiences in a year. The unit of analysis are roads in the district, identified per road by a Route ID. Using data over time, I can examine traffic data in the two years before and after the height of the pandemic in 2020. Comparing traffic patterns over time will allow me to examine changes in trends, patterns, and quantities.

I first plot the traffic patterns on map to visually examine average traffic in DC by year. I then calculate Moran’s I for the traffic of roads for each year of data to assess spatial autocorrelation of traffic. Finally, I run Monte Carlo simulations to assess the statistical significance of the Moran I values. Taken together, these methods show if spatial autocorrelation of traffic exists or has changed over time in the city, as new roads have developed traffic patterns as a result of trends caused by the COVID-19 pandemic.

Reading layer `2018_Traffic_Volume' from data source 
  `/Users/holtcochran/PPOL6805_final_project/Data/2018_Traffic_Volume.geojson' 
  using driver `GeoJSON'
Simple feature collection with 22835 features and 17 fields
Geometry type: MULTILINESTRING
Dimension:     XY, XYZ
Bounding box:  xmin: -77.11664 ymin: 38.79323 xmax: -76.90953 ymax: 38.99526
z_range:       zmin: 0 zmax: 0
Geodetic CRS:  WGS 84
Reading layer `2019_Traffic_Volume' from data source 
  `/Users/holtcochran/PPOL6805_final_project/Data/2019_Traffic_Volume.geojson' 
  using driver `GeoJSON'
replacing null geometries with empty geometries
Simple feature collection with 21434 features and 17 fields (with 45 geometries empty)
Geometry type: GEOMETRY
Dimension:     XY, XYZ
Bounding box:  xmin: -77.11664 ymin: 38.79323 xmax: -76.90953 ymax: 38.99526
z_range:       zmin: 0 zmax: 0
Geodetic CRS:  WGS 84
Reading layer `2020_Traffic_Volume' from data source 
  `/Users/holtcochran/PPOL6805_final_project/Data/2020_Traffic_Volume.geojson' 
  using driver `GeoJSON'
replacing null geometries with empty geometries
Simple feature collection with 7734 features and 17 fields (with 4 geometries empty)
Geometry type: LINESTRING
Dimension:     XY, XYZ
Bounding box:  xmin: -77.11664 ymin: 38.79323 xmax: -76.90953 ymax: 38.99236
z_range:       zmin: NA zmax: NA
Geodetic CRS:  WGS 84
Reading layer `2021_Traffic_Volume' from data source 
  `/Users/holtcochran/PPOL6805_final_project/Data/2021_Traffic_Volume.geojson' 
  using driver `GeoJSON'
replacing null geometries with empty geometries
Simple feature collection with 21702 features and 17 fields (with 170 geometries empty)
Geometry type: GEOMETRY
Dimension:     XY, XYZ
Bounding box:  xmin: -77.11664 ymin: 38.79323 xmax: -76.90953 ymax: 38.99526
z_range:       zmin: 0 zmax: 0
Geodetic CRS:  WGS 84
Reading layer `2022_Traffic_Volume' from data source 
  `/Users/holtcochran/PPOL6805_final_project/Data/2022_Traffic_Volume.geojson' 
  using driver `GeoJSON'
Simple feature collection with 8217 features and 16 fields
Geometry type: MULTILINESTRING
Dimension:     XY, XYZ
Bounding box:  xmin: -77.11664 ymin: 38.79323 xmax: -76.90953 ymax: 38.99236
z_range:       zmin: 0 zmax: 0
Geodetic CRS:  WGS 84

Exploratory Data Analysis (EDA)

I first plot the traffic data on a map to explore trends in traffic patterns at a high level. Colors of the roads indicate the amount of Average Annual Traffic by road, which darker shades of red indicating areas of higher traffic. The map is displayed by year, allowing for comparison of traffic between years.

Figure 1. Average Annual Daily Traffic (AADT) in Washington, DC

I then examine the amount of traffic at a more granular level for each year, allowing for comparisons of the percentage of high-traffic routes for each year. Figure 2 displays a histogram of AADT per year breaking down traffic into larger categories for comparison.

Figure 2.

Moran’s I

Next I calculate the Moran’s I for the Average Annual Daily Traffic for each road with data for each year from 2018 - 2022. Moran’s I statistics are used to quantify spatial correlation and assess clustering of data. I use Moran’s I to assess if traffic in DC is spatially auto correlated, and the strength of the clustering of traffic on roads. To reduce computational complexity, I filter to only assess AADT > 50,000 - this only includes roads that have average traffic of more than 50,000 drivers a day. This focuses the analysis on roads with large quantities of drivers and traffic, which is the focus of this research.

Figure 3 displays Moran’s I comparison across years. There is peak Moran’s I value of 0.84 in 2019, with a clear dip to 0.74 in 2021.

Figure 3.

Spatial Autoregressive (SAR) Model

I then a spatial autoregressive (SAR) model to test the statistical significance of the changes in dispersion of the traffic in Washingon, DC.

Discussion

Conclusion